Scribd
Upload
3 views9 pages

OLS and Fixed Effects Analysis Results

The document presents results from two hypotheses tested using OLS and Fixed Effects models on campaign and contribution datasets. For Hypothesis 1, the OLS regression indicates that campaign quality and goal positively influence collected funds, while business ventures negatively impact it. Hypothesis 2 explores the relationship between daily contributions and existing funds, showing a significant effect with clustered standard errors, but a low R-squared value suggests limited explanatory power.

Uploaded by

Albin Ting
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
3 views9 pages

OLS and Fixed Effects Analysis Results

The document presents results from two hypotheses tested using OLS and Fixed Effects models on campaign and contribution datasets. For Hypothesis 1, the OLS regression indicates that campaign quality and goal positively influence collected funds, while business ventures negatively impact it. Hypothesis 2 explores the relationship between daily contributions and existing funds, showing a significant effect with clustered standard errors, but a low R-squared value suggests limited explanatory power.

Uploaded by

Albin Ting
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Notes: official code is at the bottom.

For hypothesis 2, focus on results with clustering and fixed


effects model

Code result:
C:\Users\sfdas>python [Link]
Files loaded successfully!

--- RESULTS FOR HYPOTHESIS 1 ---


OLS Regression Results
========================================================================
======
Dep. Variable: collected_funds R-squared: 0.109
Model: OLS Adj. R-squared: 0.109
Method: Least Squares F-statistic: 203.9
Date: Thu, 25 Dec 2025 Prob (F-statistic): 8.80e-125
Time: 18:17:14 Log-Likelihood: -47469.
No. Observations: 5000 AIC: 9.495e+04
Df Residuals: 4996 BIC: 9.497e+04
Df Model: 3
Covariance Type: nonrobust
========================================================================
============
coef std err t P>|t| [0.025 0.975]
------------------------------------------------------------------------------------
Intercept 900.7696 124.427 7.239 0.000 656.837 1144.702
campaign_quality 164.0430 18.208 9.009 0.000 128.347 199.739
goal 0.1945 0.009 21.117 0.000 0.176 0.213
business_venture -320.5232 99.834 -3.211 0.001 -516.241 -124.805
========================================================================
======
Omnibus: 5446.150 Durbin-Watson: 2.015
Prob(Omnibus): 0.000 Jarque-Bera (JB): 1131861.574
Skew: 5.138 Prob(JB): 0.00
Kurtosis: 75.989 Cond. No. 2.32e+04
========================================================================
======

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The condition number is large, 2.32e+04. This might indicate that there are
strong multicollinearity or other numerical problems.

--- RESULTS FOR HYPOTHESIS 2 ---


OLS Regression Results
========================================================================
======
Dep. Variable: dailycontrib R-squared: 0.028
Model: OLS Adj. R-squared: 0.027
Method: Least Squares F-statistic: 113.4
Date: Thu, 25 Dec 2025 Prob (F-statistic): 2.75e-49
Time: 18:17:14 Log-Likelihood: -21077.
No. Observations: 8000 AIC: 4.216e+04
Df Residuals: 7997 BIC: 4.218e+04
Df Model: 2
Covariance Type: nonrobust
========================================================================
============
coef std err t P>|t| [0.025 0.975]
------------------------------------------------------------------------------------
Intercept 1.2738 0.078 16.346 0.000 1.121 1.427
daily_total_perk 0.0002 1.87e-05 9.380 0.000 0.000 0.000
day -0.0447 0.003 -13.405 0.000 -0.051 -0.038
========================================================================
======
Omnibus: 25270.399 Durbin-Watson: 1.082
Prob(Omnibus): 0.000 Jarque-Bera (JB): 3246414474.500
Skew: 50.190 Prob(JB): 0.00
Kurtosis: 3122.161 Cond. No. 5.16e+03
========================================================================
======

Notes:
[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.
[2] The condition number is large, 5.16e+03. This might indicate that there are
strong multicollinearity or other numerical problems.
Code:
import pandas as pd
import [Link] as smf
import [Link] as sm
import [Link] as plt

# 1. LOAD THE DATASETS


# Replace the names inside quotes with your actual file names
try:
df_campaign = pd.read_csv('BE603_campaigns_36.csv')
df_contrib = pd.read_csv('BE603_contrib_36.csv')
print("Files loaded successfully!")
except FileNotFoundError:
print("Error: The CSV files were not found. Check your file names and
folder path.")

# --- HYPOTHESIS 1 (Campaign Dataset) ---


# Testing if quality signals (images/videos) affect total funds.
# Equation: collected_funds = B0 + B1(campaign_quality) + B2(goal) +
B3(business_venture)
model1 = [Link](formula='collected_funds ~ campaign_quality + goal +
business_venture',
data=df_campaign).fit()

print("\n--- RESULTS FOR HYPOTHESIS 1 ---")


print([Link]())

# --- HYPOTHESIS 2 (Contribution Dataset) ---


# Testing for "Social Proof" (Momentum): Do existing funds attract more
daily donors?
# Equation: dailycontrib = B0 + B1(daily_total_perk) + B2(day)
model2 = [Link](formula='dailycontrib ~ daily_total_perk + day',
data=df_contrib).fit()

print("\n--- RESULTS FOR HYPOTHESIS 2 ---")


print([Link]())

# --- STEP 4: MODEL ADEQUACY (Residual Check) ---


# This creates a Q-Q plot to see if your errors are normally distributed
fig = [Link]([Link], line='s')
[Link]("Normal Q-Q Plot of Residuals (H1)")
[Link]()
Fixed effect model results: (no clustering)
--- FIXED EFFECTS RESULTS FOR HYPOTHESIS 2 ---
PanelOLS Estimation Summary
========================================================================
========
Dep. Variable: dailycontrib R-squared: 0.0081
Estimator: PanelOLS R-squared (Between): -1.3870
No. Observations: 8000 R-squared (Within): 0.0081
Date: Fri, Dec 26 2025 R-squared (Overall): -0.1249
Time: 14:46:15 Log-likelihood -2.088e+04
Cov. Estimator: Unadjusted
F-statistic: 63.594
Entities: 200 P-value 0.0000
Avg Obs: 40.000 Distribution: F(1,7799)
Min Obs: 40.000
Max Obs: 40.000 F-statistic (robust): 63.594
P-value 0.0000
Time periods: 40 Distribution: F(1,7799)
Avg Obs: 200.00
Min Obs: 200.00
Max Obs: 200.00

Parameter Estimates
========================================================================
============
Parameter Std. Err. T-stat P-value Lower CI Upper CI
------------------------------------------------------------------------------------
daily_total_perk -0.0003 3.982e-05 -7.9746 0.0000 -0.0004 -0.0002
========================================================================
============

F-test for Poolability: 2.9173


P-value: 0.0000
Distribution: F(199,7799)

Included effects: Entity


Hypothesis 2 code without clustered standard errors
# 1. LOAD THE DATASET
try:
df_contrib = pd.read_csv('BE603_contrib_36.csv')
print("File loaded successfully!")
except FileNotFoundError:
print("Error: Check your file path or name.")

# 2. PREPARE DATA FOR FIXED EFFECTS


# We use 'id' as the entity index and 'day' as the time index
df_fe = df_contrib.set_index(['id', 'day'])

# 3. RUN FIXED EFFECTS (FE) MODEL


# EntityEffects=True tells Python to compare each campaign only to itself
over time
model_fe = PanelOLS.from_formula(
'dailycontrib ~ daily_total_perk + EntityEffects',
data=df_fe
).fit()

print("\n--- FIXED EFFECTS RESULTS FOR HYPOTHESIS 2 ---")


print(model_fe.summary)
Fixed effect model results with clustering
--- FIXED EFFECTS RESULTS (CLUSTERED SE) ---
PanelOLS Estimation Summary
========================================================================
========
Dep. Variable: dailycontrib R-squared: 0.0081
Estimator: PanelOLS R-squared (Between): -1.3870
No. Observations: 8000 R-squared (Within): 0.0081
Date: Sat, Dec 27 2025 R-squared (Overall): -0.1249
Time: 12:33:08 Log-likelihood -2.088e+04
Cov. Estimator: Clustered
F-statistic: 63.594
Entities: 200 P-value 0.0000
Avg Obs: 40.000 Distribution: F(1,7799)
Min Obs: 40.000
Max Obs: 40.000 F-statistic (robust): 3.2595
P-value 0.0710
Time periods: 40 Distribution: F(1,7799)
Avg Obs: 200.00
Min Obs: 200.00
Max Obs: 200.00

Parameter Estimates
========================================================================
============
Parameter Std. Err. T-stat P-value Lower CI Upper CI
------------------------------------------------------------------------------------
daily_total_perk -0.0003 0.0002 -1.8054 0.0710 -0.0007 2.724e-05
========================================================================
============

F-test for Poolability: 2.9173


P-value: 0.0000
Distribution: F(199,7799)

Included effects: Entity


Official code:​
import pandas as pd
import [Link] as smf
import [Link] as sm
import [Link] as plt
from [Link] import PanelOLS

# 1. LOAD THE DATASETS


try:
df_campaign = pd.read_csv('BE603_campaigns_36.csv')
df_contrib = pd.read_csv('BE603_contrib_36.csv')
print("Files loaded successfully!")
except FileNotFoundError:
print("Error: The CSV files were not found. Check your file names.")

# --- STEP 3: RUNNING THE REGRESSIONS ---

# --- HYPOTHESIS 1 (Cross-Sectional OLS) ---


# Testing if quality signals affect total funds using the Campaign
dataset.
model1 = [Link](formula='collected_funds ~ campaign_quality + goal +
business_venture',
data=df_campaign).fit()

print("\n" + "="*40)
print("--- RESULTS FOR HYPOTHESIS 1 (OLS) ---")
print("="*40)
print([Link]())

# --- HYPOTHESIS 2 (Fixed Effects Panel Model) ---


# Testing for Social Proof using the Contribution dataset.
# We prepare the data by setting the MultiIndex (Entity=id, Time=day)
df_fe = df_contrib.set_index(['id', 'day'])

# We use Clustered Standard Errors to address autocorrelation


model2_fe = PanelOLS.from_formula(
'dailycontrib ~ daily_total_perk + EntityEffects',
data=df_fe
).fit(cov_type='clustered', cluster_entity=True)

print("\n" + "="*40)
print("--- RESULTS FOR HYPOTHESIS 2 (FIXED EFFECTS + CLUSTERED SE) ---")
print("="*40)
print(model2_fe.summary)

# --- STEP 4: MODEL ADEQUACY (Residual Check) ---


# Creating a Q-Q plot for Hypothesis 1 to check for normality/outliers
fig = [Link]([Link], line='s')
[Link]("Normal Q-Q Plot of Residuals (Hypothesis 1)")
[Link]()

# Optional: Print Durbin-Watson for Hypothesis 1 to discuss


autocorrelation
from [Link] import durbin_watson
dw1 = durbin_watson([Link])
print(f"\nDurbin-Watson for H1: {dw1:.4f}")

You might also like